Agentic AI is only as effective as the data it can securely access, understand, and act upon. Yet most enterprise applications were never designed for autonomous agents to safely discover information, execute queries, or interact with business systems. As organizations move from chatbots to AI agents, data architecture suddenly becomes a critical design concern. In this session, we’ll explore how to transform existing enterprise data into secure, agent-ready capabilities using modern retrieval patterns and agent protocols. Through a live end-to-end demo, we’ll connect an agentic orchestration built with the Microsoft Agent Framework to enterprise data stored in Azure SQL and Cosmos DB. You’ll see how to structure, categorize, and expose application data so AI agents can retrieve knowledge, execute governed operations, and reason over live business data while respecting permissions, boundaries, and context. The session demonstrates two complementary approaches to agentic data access: Retrieval-Augmented Generation (RAG) for semantic knowledge retrieval and grounding; and Model Context Protocol (MCP) for structured, tool-based access to operational systems. We’ll compare when to use RAG versus MCP, how those choices influence your overall data architecture, and the trade-offs between vectorized retrieval and direct operational querying. We’ll also cover the security implications of agentic systems, including which classes of agent actions should be avoided entirely in enterprise environments. By the end of the session, you’ll understand how to: 1. Prepare enterprise data for agentic workloads; 2. Design secure boundaries for AI agent access; 3. Combine relational and document databases in AI-driven workflows; 4. Build reusable MCP tools on top of existing systems; 5. Avoid common pitfalls such as overexposed schemas, hallucinated queries, excessive permissions, and unsafe autonomous actions.